[Zope3-Users] Zemantic 0.5 released

[Michel Pelletier]

While the majority of the Paris sprinters were working on the 2.8/Five 
integration, Tres Seaver, Tarek Zaide and I worked to improve Zemantic.  I am 
very happy to annouce the 0.5 release based on this great week of work can be 
found on the zemantic web site:

http://zemantic.org/

Zemantic is a semantic web catalog for Zope 3.  It takes RDF data from any 
source, local or remote, linearizes it into "statments" (also called 
"triples"), and indexes those statements in three dimensions, so that any 
subject, predicate or object in the statements can be quickly searched for.  
More information and links to semantic web technology can be found on the 
zemantic web page.

At the sprint our focus was to improve access for other developers to use 
Zemantic.  We set out from the beginning therefore to integrate Zemantic into 
a real-world application.  The application we chose was CPS web mail, which 
is also based on Five (and now, Zope 2.8).   I am happy to report that our 
experiment was  a total success, CPS web mail now uses Zemantic as its core 
search engine, completely replacing the ZCatalog.

Some of the new things in Zemantic are:

  - Full text indexing of RDF literals.  While this is not the final word by 
any means on indexing RDF in different ways, it's a great start.

  - More tests that are much better at flexing the core of Zemantic.  Tres and 
Tarek in particular wrote many tests.

  - Tres spend the first day of the sprint making Zemantic a disutils package, 
making installation much easier.

  - Some documentation improvements were made, although there is still a lot 
of room for improvement.

Some future goals we brainstormed at the sprint:

  - Pluggable query languages: interfaces to plug the many RDF query languages 
into Zemantic.  For starters, there is currently a Sparql implementation 
written in Python that I am looking at integrating.

  - More configurable indexing.  In addition to indexing the RDF graph 
structure,  Zemantic indexes all text literals, but this is possibly overkill 
and not very configurable.   Zemantic also applies no interpretation to the 
RDF data, so it cannot interpret (and therefore index specially) object value 
data like dates.  A more flexible index layer on top of Zemantic could be 
used to specify special indexing, probably based on the predicate or some 
other kind of pluggable rules.

  - Integrating Zemantic with natural language processors like NLTK, so that 
RDF data can be extracted from unstructured textual documents.

This is only the beginning, I have great confidence in both Zope 3 and the 
semantic web, and I think that 2005 will be the year of both.  The Paris 
sprint has been a great success, and I'd really like to thank everyone who 
participated and particularly Nuxeo for making it happen!

-Michel